Uploaded image for project: 'JDK'
  1. JDK
  2. JDK-8099421

Improve startup time on Windows with 6u10 by not initializing Nimbus Look and Feel

    XMLWordPrintable

    Details

    • Type: Enhancement
    • Status: Resolved
    • Priority: P3
    • Resolution: Fixed
    • Affects Version/s: fx1.0
    • Fix Version/s: fx1.2
    • Component/s: javafx
    • Environment:

      windows, 6u10

      Description

      Tests show that not initializing Nimbus look and feel improves the startup time of FXFramer application (simplest desktop FX app) by about 10%.

      Here's the data provided by David Dagastine - testing Igor's fix which just disabled the initialization of Nimbus L&F:

      -----------
      Here's the results compared to b666 and b773 on Windows Vista. Very nice improvement of nearly 12% over b773, bringing startup time on vista down to 0.85 sec. Richard, could you point me to jars based on b773 with the NoLookAndFeel change? I'd like to compare the performance impact. I'm running XP tests now.

      ==============================================================================
      startup3fx-b666-search.1:
       Benchmark Samples Mean Stdev Geomean Weight
       startup3 30 0.77 0.21
         Noop 30 0.54 0.35 0.50
         XFramer 30 1.16 0.01 0.50
      ==============================================================================
      startup3fx-b773-search.2:
       Benchmark Samples Mean Stdev %Diff P Significant
       startup3 30 0.62 0.11 18.65 0.002 Yes
         Noop 30 0.41 0.18 23.66 0.084 *
         XFramer 30 0.96 0.04 16.91 0.000 Yes
      ==============================================================================
      startup3fx-b773-dev-search.1:
       Benchmark Samples Mean Stdev %Diff P Significant
       startup3 30 0.55 0.10 28.28 0.000 Yes
         Noop 30 0.37 0.19 32.27 0.021 *
         XFramer 30 0.85 0.02 26.71 0.000 Yes
      ==============================================================================
       * - Not Significant: A non-zero %Diff for the mean could be noise. If the
           %Diff is 0, an actual difference may still exist. In either case, more
           samples would be needed to detect an actual difference in sample means.


      Using b773 as a baseline:
      ==============================================================================
      startup3fx-b773-search.2:
       Benchmark Samples Mean Stdev Geomean Weight
       startup3 30 0.62 0.11
         Noop 30 0.41 0.18 0.50
         XFramer 30 0.96 0.04 0.50
      ==============================================================================
      startup3fx-b773-dev-search.1:
       Benchmark Samples Mean Stdev %Diff P Significant
       startup3 30 0.55 0.10 11.83 0.010 Yes
         Noop 30 0.37 0.19 11.27 0.341 *
         XFramer 30 0.85 0.02 11.79 0.000 Yes
      ==============================================================================
       * - Not Significant: A non-zero %Diff for the mean could be noise. If the
           %Diff is 0, an actual difference may still exist. In either case, more
           samples would be needed to detect an actual difference in sample means.

        Attachments

          Issue Links

            Activity

              People

              Assignee:
              rbair Richard Bair (Inactive)
              Reporter:
              tdv Dmitri Trembovetski (Inactive)
              Votes:
              0 Vote for this issue
              Watchers:
              7 Start watching this issue

                Dates

                Created:
                Updated:
                Resolved:
                Imported: